The type of nerd that watches this content is more than capable of groking long form videos. I hunger for this deep and hard earned knowledge. It saves me hundreds of hours of trial and error.
This is flat out awesome. This topic is full of concepts that are such a challenge to grasp even after 'reading' a few high-speed design books. In general, the books dive straight into the math without a general overview. Thank you for taking the time to present this!
I wish I could animate what I see in my head when I look at high speed stuff. Magnetic fields look to me like the trails the bullets leave on the Matrix movies. (helps with the orientation along the direction of travel)
I wish there was a comprehensive video series on modern PCB design on UA-cam. Right from baby steps to a finished professional product. The student can then do their own research regarding specialization into a particular domain, whether power supplies, mixed signal design etc
Check out Robert Fedevel. While his stuff is Altium based, most of the information he shares is applicable anywhere. His older, longer uploads and the ones about redesigning an arduino uno are super helpful. His series of design tips also covers lots of simple aspects that are very helpful to both the hobbyist and student. Dave has a few uploads on the EEVBlog and EEVBlog2 that both explain the design process and a full timelapse of the layouts. Watch those to get a better understanding of how to layout a board in blocks. I'm a big advocate for learning with KiCAD as it is unlimited, very powerful, runs on legacy hardware, and is free. KiCAD 5 was released a few months ago and has proven to be a big step forward in the maturity of the software. The channel Contextual Electronics does lots of uploads and tutorials explaining KiCAD features and how to use them. After gaining some basic skills, KiCAD also supports lots of user mods and plugin features. The KiCAD forum is very active and helpful as well. I also save everything I come across, watch, and find helpful. Quite awhile back I decided to make all of these references public along with all of the potential references I have on this account. Contrary to the appearance of my name and thumbnail, I'm not monetized, and couldn't care less about self promo. I'm just a dumb hobbyist, but my playlists contain tons of useful electronics specific references that are indexed alphabetically (although other users are not able to sort alphabetically as they appear to me). My subscriptions are also publicly listed. I watch a bunch of edutainment channels you'd have to filter through, but I subscribe to everything electronics related that I come across from small channels and users to manufacturers. Check out my "Intermediate PCB Design" playlist for references. I also have a bunch of playlists covering all aspects of DIY PCB fabrication. Even with the cheap board houses available, I learned quite a bit from etching my own mistakes first. GL ;) -Jake
A I, could you kindly elaborate on what you mean by specialization? What is there to specialize when doing PCB design for power supplies e.t.c? I mean there are just a few guidelines to remember and most PCBs will have digital & analogue processing circuits along with power supply portion.
@@UpcycleElectronics Thank you ! I will surely check out the stuff that you've listed. I thought the good stuff from Robert Fedeval was paid courses ??? I guess one can find good stuff on his youtube channel too. I usually work with SMPS and have to get my boards designed by another guy. While I know what specification I should give the PCB designer, I feel I would save a lot of time learning and doing the stuff myself. Lets see how the learning process goes.
@@優さん-n7m Well power supplies, switchmode ones, usually have high frequencies on the board so that may actually change the trace impedance characteristics. Plus you have to consider things like isolation, emi and various compliance constraints according to the code for that particular power supply. Similarly, I'm sure the design of a purely analog board with just DC around the components would be different from a board that would have digital signals running around alongwith the analog ones. Thats what I meant by specialization.
I don't get the horizonal vs vertical impedance difference. Isn't the difference the exact *location* on the horizontal or vertical plane (i.e. either over the glass or epoxy) rather than actually being horizontal or vertical? The weave looks consistent in either direction if you have the trace over the actual glass weave.
You're correct! Altera has a wonderfully documented look at the fiber weave effect (just Google "Altera Fiber Weave Effect") that not only explains the effect in depth, it also shows some great ways to compensate for it (like keeping traces at a 10° angle)
Yes, correct. It is all about the type of material your have around your trace, glass or epoxy. If the prepreg is uniform there should be no difference between the horizontal and vertical trace given the trace is located at a similar location relative to the weave.
Yes, I was about to post the same thing. It's much worse than a difference between horizontal and vertical; you can have a difference between two horizontal traces, running side by side. If he got a difference between horizontal and vertical, that was just because they lined up differently with the weave.
They are actually rectangles if you look closely, and there is a difference whether a trace falls right on a horizontal strand or on the interleaved gaps just next to it. Offsetting the trace or angling the PCB statistically lowers the issues but can still hit it "just wrong". The best answer is to roll the weave flat during fab but it is more expensive. www.intel.com/content/dam/altera-www/global/en_US/documentation/bib1485555122987/osg1487021808536.jpg
That's a great app note but yet there are things in it I don't agree with. Let Ritchey (Right the First Time) recently told a bunch of us that he talked with some of the people that wrote these app notes and asked where they get their information from, and the answer was usually that they just copied it from the previous app note, back to the beginning of app notes in some cases. For example, doe sit REALLY make sense to put an inductor in the form of a ferrite bead in series with a power supply when the goal is to lower inductance, not raise it? There is another myth that got it's genesis in something wrong in an early ECL design guide, and has been with us ever since.
Too short! This could have been 30m longer and I'd still watch it to the end. There is a huge shortage of tech lecturers online and Bil is just a treat to listen to.
Having migrated to FPGA development for high-speed buses (currently doing a Cyclone V with USB 3 transceiver), its amazing how good looking layouts and routing immediately becomes so critical.
Suddenly, it made sense. The return path follows the signal path because of capacitance, mostly, but also inductance. It follows the path of least impedance, although it doesn't match the one of least resistance. Thanks for the detailed video, and for the aha moment, Bill!
In-depth definitely yes. But I have some constraints - I don’t start viewing videos that are 5 hours long. In fact, I seldom pick a video that is over 1 hour. Sometimes I have a time slot of just 10 or 15 minutes available and I may pick a video that fits in that window. But generally my observation is that even in-depth videos fit in one our. I mean, some videos seem to have a repetitious narration “be extremely careful” and similar repeated ad nauseum. That is not in-depth as far as I am concerned.
You can always watch part of a longer video, and then come back to it at a later point if you dont have time to watch a longer video start to finish in one go? I have personally done this, no it's not ideal. Maybe long format videos are not what hackaday want to produce, which I think is a shame when people like Bil have such an amazing amount of knowledge and stories they can share. Sure I can go and do my own research to fill in the gaps, but then you miss out on the personality of the presenter.
BilHerd As I thought I was expressing, in my case it is more a question of available time windows than attention span. On the other hand, splitting long ones is not really working either. If I close my watching - and my iPad -. Next time I open UA-cam, the interrupted story is nowhere to be seen. It may return to the “Recommended for you” offerings maybe in a week or maybe in a month. I guess I could mark it as one I want to watch later, but I have not got to that kind of habit.
Thanks for the high speed ride on this Mr. Herd. Definitely needs more time on this subject... it's been a while since I did this stuff in university. BTW - I picked up one of your C128's recently (a C128D). Looking forward to poking around in it and see how the Z80 and 68xx play with one another. Cheers, - Eddy
Great video, thanks! I had my first experience of highish speed layout around 5 years ago when I started using TFTs which necessitated some SDRAM. I read all the app notes with their rules of thumb and came to the conclusion it was impossible if you stuck to the rules and wanted a 4 layer board. But as has happened so many times in my career I did my best and it was totally fine. (Measure twice, cut once.)
Oh I always wanted to know about the effect of vias. Turns out (at least done as stubs) they cause a way worse effect than right angles. (well yeah I get that its an apples to oranges comparison but people seem to get a lot more bothered about right angles than via placement in traces).
I never thought that the issues with square-corner traces would come up at so high frequencies, I just assumed that the reflections of the propagating waves would produce EMC problems at any frequency haha! Good to know, because it helps you reduce the PCB size and the traces routing. Great vid!
Somewhere in the Multiverse there's a version of me who finished his Science degree and is going "Duh!" after watching this. My version in this universe is just amazed at all the magic. Great video Bil! Thanks for posting.
I was taught to avoid sharp corners because the etch could undercut the resist at those sharp features, reducing the trace width, sometimes etching through the trace. Still a good reason to avoid sharp corners.
I think its the attention span of the average Hackaday viewer that we most want to accommodate. My editor would probably like it if I made more and shorter videos, :) There is just so much post-editing time involved in making a video that I try and cram as much into one as I can.
He didn't mention it, it is related to the reflected wave ratio, here is a tektronix link that can help to understand www.google.com/url?sa=t&source=web&rct=j&url=www.tek.com/document/fact-sheet/tdr-impedance-measurements-foundation-signal-integrity&ved=2ahUKEwiIvtX33uXnAhUEnZ4KHS_oDRMQFjADegQIAhAB&usg=AOvVaw2NZJP5Oh8zyAC92obfjuq7
Thanks Bil and Hackaday! This subject is a bit over my head, although it's one I really want to learn. I wish you would consider doing more of an abstract overview that covers high speed design 101 from the worst case scenario to radio. I'm a bit hazy about when and how high speed design comes into play. I wish you would assume the average hobbyist starts with a single sided home etched design, then graduates to a double sided, then further specialties like impedance controlled copper clad, or board houses. These days, IMO, the path of least resistance is to start with AVR's before radio. I don't know the limitations of each of these applications and will inevitably learn the hard way,.. unless lady luck (google) leads me to the right references. I know Alan W. has a bunch of stuff about doing Manhattan Style boards and dead bug/point to point designs. I just haven't played with much radio stuff yet. I don't know how that fits into the possibilities between single and double sided boards. I know I've taken apart lots of consumer products like TV's and Radios that were built on single sided boards w/jumpers. I don't know if the EEs have Tinkerbell on their shoulder adding pixie dust into their designs or if single sided boards are far more practical than they appear due to all the information people share about more advanced designs. Such info would be extremely helpful to me as a dumb hobbyist blindly wronging my way to right ;) -Jake
@@優さん-n7m In my uneducated sense of board design I think of "radio stuff" as anything around or above the UHF frequency range. I think the biggest hobbyist related application of, and introduction to, such things is amateur radio. I am probably mistaken. I just mentally group all of this in a figurative black box labeled magic... for now.
Jake, There’s more to say on this subject than can fit into a UA-cam comment, but suffice to say you’re actually not far off in saying that it’s ‘magic’ once you get past a certain speeds. In fact one of the most popular books on the subject of high speed design joking calls it “black magic” (www.amazon.com/High-Speed-Digital-Design-Handbook/dp/0133957241). It certainly feels that way to most of us, even if you understand the theory. One thing that’s helpful to remember is that the rules haven’t changed (the math and physics are the same), but lots of things you can safely ignore at low speeds become headaches and other things that seem significant at low speeds may not be as important. So you’ll need to learn new set of rules of thumb and expectations at those speeds, or better yet will need a strong intuitive understanding of the underlying math and physics. Simply put: the faster the speed, the more the geometry of your traces matter. If you’re interested in working with Bluetooth, FPGAs, PCIe, WiFi, etc it’s going to be difficult to get good results with home-etched boards and rules of thumb. If you’re working with microcontrollers and I2C and SPI etc over short distances, you won’t need to worry about all of this to get it to work, but if you understand it, it will work better. If you want to learn more, I the above textbook and Eric Bogatin (mentioned in the video: www.bethesignal.com/bogatin/) are great resources to get started. If that’s more than you want to take on now, I highly recommend ‘The Art of Electronics’, if you haven’t read it yet (www.amazon.com/gp/aw/d/0521809266/). It covers lots of key topics in a very approachable way that will help build you up for more of the high speed theory. From my perspective, if you’re able to do everything you want to do right now with electronics, don’t worry about it, but if you’re running into limitations with your data speeds or want to do wireless transmission without buying a pre-built module, then start small, make mistakes you can learn from, and find someone smart you who will let you ask dumb questions.
@@michaelkling4994 Thanks for the book references. I'm not having too many problems yet, but I've only done around a dozen or so prototype designs so far with my biggest being a simple ATmega644 AVR Transistor Tester dev board, so nothing serious yet. I have The Art of Electronics 2nd edition along with a few other books. I generally like to get an abstract idea of what to be on the lookout for, that's all. I like to build up a map in my mind of how things I don't really understand yet connect to what I'm doing or want to do in the near future. I'm disabled and have all the time in the world. I'm just really really slow at getting projects done and I don't get out much. So...I'm usually asking lots of dumb questions in YT comments ;) It's all in good fun. -Jake
Upcycle Electronics, yeah I totally get what you’re saying about having a mental map. That’s part of why I got my EE degree: I could conceptually understand how you could build a circuit out of wires and chips, and how you could code a piece of software, but things got pretty hazy for me in-between. What I ended up finding out the hard way, was that since electricity is (usually) invisible, the only way to describe what is happening and how it works is through math. I’m certainly not a math genius, but I found that as a I fought through it, math helped provide the language and framework to open up semiconductors and electromagnetism and all the other concepts I needed to fill in the gaps in my mental model. One of my biggest breakthroughs was discovering frequency-domain mathematics (eg Fourier transforms). It took some time but once I started conceptualizing all signals as collections of sine waves at different frequencies and amplitudes, it opened up all kinds of things, particularly in filters, signal processing, high speed electronics, and electromagnetism. If you haven’t gone down the frequency-domain rabbit hole yet, I bet it would be well worth you time to spend some time looking up Fourier series and transforms. The overall concept is more important than the details of the math for day to day use, and it helps bridge the gap between what you’re working with now and higher speed boards and projects. Best of luck to you :-)
Would be nice to see a video that covered possible bus speeds of the classic era of 33 Mhz on up into the Ghz range and sample common speeds along the way (for data transfers). Then do a sweep with signals that simulate actual RF (usually based around sine waves) covering HF -> VHF and maybe UHF. I was curious to see some effects on the different types of signal because of the potential for ringing (which you did show some ripple from the vias) Thanks for the great video!
With the right angle traces, your board actually exaggerates the problem a little bit, because each turn of your square snake allows a portion of the trace to couple to the previous turn. Your typical right angle trace that we don't have to worry about will not have geometry that allows the signal trace to couple to itself like that. This just further proves how pointless it is to remove right angles until you get to the GHz-ish ranges.
I would want to see the right angle thing done with just one or two right angles, where the path then really turns (or is Z shaped). I think in that configuration you might have quite a bit of capacitive coupling between the traces, that is I wouldn't be surprised if the picture looked similar had you left the "right angle ends" off the lines.
:) That's right Maxwell's equations only speak to Electric (E) and Magnetic (B) fields, Current (J), and the speed of light (c). No voltage directly, that's the potential in the E field.
Thank you so much for the video, I had a question then, next year I have to choose my major for college and I was wondering to learn these things what major should I choose, is that electrical engineering! I know that my friend is studying electrical but they do not learn about these things, to learn these should they take their master as well or is there any specific major, like electronics engineering?
The post that accompanies this video can be seen at hackaday.com/2019/01/24/video-putting-high-speed-pcb-design-to-the-test/ There are additional graphics there that support what go flying by in the video.
Re: ground planes. The graphic you show at 5:12 is even true at hundreds of kHz. Current has little to do with electron flow which is centimeters per hour, yet that's still how we're all brainwashed in University and by KCL circuit simulators. Only Faraday's and Maxwell's equations correctly explain current density at frequencies above 100kHz. Good luck with those KCL spice simulators. Kudos for the graphic at 5:12......I can guarantee you most ee graduates today will never understand that.
Weave anisotropy induces mode conversion primarily, due to permittivity variation/intra differential skew. The impedance impact is negligible however, because impedance has more to do with the geometry of the copper conductors. Also, 106 glass isn't used by anyone making a high-speed board. It's horrible. A board house wouldn't even allow it. It would be Megtron 6 within 3313 glass
That was really nice! It would be great if you could talk more about what frequency ranges should be concerned with the different techniques e.g. do right angles matter at 320MHz, 3.2GHz, 10GHz, etc.. also what is required for the really high speed links available in recent Xilinx FPGAs like 10Gb/s, 16Gb/s, 25Gb/s
Right angles to matter since at right angles the track width is larger and that impacts the characteristic imedance. Now the design might work just fine, or this impedance mismatch might ruin the design completely. It depends on the frequency of the signal.
Mittered bend is the best. Statement, that 90° corner bends are OK for PCB design at high frequencies is simply not true. For frequencies around 80GHz, difference in S21 can be as much as 0.4dB between mittered and 90° corner bend and return loss difference around 14dB! I simulated that in ADS for 101.6um thick laminate with Er=3.66 and for 211um wide microstrip line. I bet that this would translate into real world as well. No RF engineer would ever design with 90° corner bends.
@@BilHerd LOL, sure 'nuff. May I ask what the other two volumes next to it are? ...and any opinions on Ott's title, "Electromagnetic Compatibility Engineering"?
@@Peter_S_ They are 'Noise Reduction Techniques in Electronic Systems' and the Motorola 'MECL System Design Handbook'. He talks about them in his other PCB design video.
This stuff almost killed all the fun of design for me. When you have over 10 thousand digital signals, all with picosecond rise/fall times you spend all your time on routing and analysis instead of design. Automated tools for analysis helped but it still took forever. We lost a few guys just out of frustration. Interesting physics but that can carry your interest only so far.
Thanks for the video but can you please get a better sound setup? The white noise in the background is super annoying. As they say, the cobbler always wears the worst shoes.
It’s still basically magic, we know all about electrodynamics and quantum-electrodynamics to very very high precision, but nobody can actually solve the equations for real world applications, the math is just too complex on the macro scale...
Legally change your name or forge the credentials and then create a new google user. If caught the google police access your personal profile and it's not pretty, your GPS stops working, your credit rating tanks, and you can only watch kitten video's on youtube.
Let's think about the public's reaction to this video... Like... WHAT? So many errors in content, labeling, units (both time and frequency) that this video needed MASSIVE edits to make it understandable, let alone usable.
Very interesting -- thanks! For anyone else who doesn't know, Bill Herd is a design-engineering legend; look at the Wikipedia article for more info: en.wikipedia.org/wiki/Bil_Herd
FYI "High Speed Design" (for PCB's) has historically referred to routing for high frequency on PCB's. You could search for "Fast Turnaround" design or "Fast Layout Techniques" if you are looking for that kind of info. Cheers, - Eddy
What matters more than any of this is the ground under these traces. Zero discussion in this video. What is he testing? Can't be solid ground plane he has vertical trace on backside. No ground at all? A ring around the edges grounding the SMA? Pretty poor. I'd rather see video showing gaps in ground plane and cross wires underneath.
I thought the video would be about quickly making a PCB from circuit design to completed board in just a few days. Disappointed now. Unsubscribing! /sarc :-o
The type of nerd that watches this content is more than capable of groking long form videos. I hunger for this deep and hard earned knowledge. It saves me hundreds of hours of trial and error.
This is flat out awesome. This topic is full of concepts that are such a challenge to grasp even after 'reading' a few high-speed design books. In general, the books dive straight into the math without a general overview.
Thank you for taking the time to present this!
I wish I could animate what I see in my head when I look at high speed stuff. Magnetic fields look to me like the trails the bullets leave on the Matrix movies. (helps with the orientation along the direction of travel)
I wish there was a comprehensive video series on modern PCB design on UA-cam. Right from baby steps to a finished professional product. The student can then do their own research regarding specialization into a particular domain, whether power supplies, mixed signal design etc
Check out Robert Fedevel. While his stuff is Altium based, most of the information he shares is applicable anywhere. His older, longer uploads and the ones about redesigning an arduino uno are super helpful. His series of design tips also covers lots of simple aspects that are very helpful to both the hobbyist and student.
Dave has a few uploads on the EEVBlog and EEVBlog2 that both explain the design process and a full timelapse of the layouts. Watch those to get a better understanding of how to layout a board in blocks.
I'm a big advocate for learning with KiCAD as it is unlimited, very powerful, runs on legacy hardware, and is free. KiCAD 5 was released a few months ago and has proven to be a big step forward in the maturity of the software. The channel Contextual Electronics does lots of uploads and tutorials explaining KiCAD features and how to use them. After gaining some basic skills, KiCAD also supports lots of user mods and plugin features. The KiCAD forum is very active and helpful as well.
I also save everything I come across, watch, and find helpful. Quite awhile back I decided to make all of these references public along with all of the potential references I have on this account.
Contrary to the appearance of my name and thumbnail, I'm not monetized, and couldn't care less about self promo. I'm just a dumb hobbyist, but my playlists contain tons of useful electronics specific references that are indexed alphabetically (although other users are not able to sort alphabetically as they appear to me). My subscriptions are also publicly listed. I watch a bunch of edutainment channels you'd have to filter through, but I subscribe to everything electronics related that I come across from small channels and users to manufacturers.
Check out my "Intermediate PCB Design" playlist for references. I also have a bunch of playlists covering all aspects of DIY PCB fabrication. Even with the cheap board houses available, I learned quite a bit from etching my own mistakes first.
GL ;)
-Jake
@@UpcycleElectronics
Can KiCAD do things like simulation of a DDR3 interface to evaluate the PCB design?
A I, could you kindly elaborate on what you mean by specialization? What is there to specialize when doing PCB design for power supplies e.t.c? I mean there are just a few guidelines to remember and most PCBs will have digital & analogue processing circuits along with power supply portion.
@@UpcycleElectronics Thank you ! I will surely check out the stuff that you've listed. I thought the good stuff from Robert Fedeval was paid courses ??? I guess one can find good stuff on his youtube channel too. I usually work with SMPS and have to get my boards designed by another guy. While I know what specification I should give the PCB designer, I feel I would save a lot of time learning and doing the stuff myself. Lets see how the learning process goes.
@@優さん-n7m Well power supplies, switchmode ones, usually have high frequencies on the board so that may actually change the trace impedance characteristics. Plus you have to consider things like isolation, emi and various compliance constraints according to the code for that particular power supply. Similarly, I'm sure the design of a purely analog board with just DC around the components would be different from a board that would have digital signals running around alongwith the analog ones. Thats what I meant by specialization.
I don't get the horizonal vs vertical impedance difference. Isn't the difference the exact *location* on the horizontal or vertical plane (i.e. either over the glass or epoxy) rather than actually being horizontal or vertical? The weave looks consistent in either direction if you have the trace over the actual glass weave.
You're correct! Altera has a wonderfully documented look at the fiber weave effect (just Google "Altera Fiber Weave Effect") that not only explains the effect in depth, it also shows some great ways to compensate for it (like keeping traces at a 10° angle)
Yes, correct. It is all about the type of material your have around your trace, glass or epoxy. If the prepreg is uniform there should be no difference between the horizontal and vertical trace given the trace is located at a similar location relative to the weave.
Yes, I was about to post the same thing. It's much worse than a difference between horizontal and vertical; you can have a difference between two horizontal traces, running side by side. If he got a difference between horizontal and vertical, that was just because they lined up differently with the weave.
They are actually rectangles if you look closely, and there is a difference whether a trace falls right on a horizontal strand or on the interleaved gaps just next to it. Offsetting the trace or angling the PCB statistically lowers the issues but can still hit it "just wrong". The best answer is to roll the weave flat during fab but it is more expensive.
www.intel.com/content/dam/altera-www/global/en_US/documentation/bib1485555122987/osg1487021808536.jpg
That's a great app note but yet there are things in it I don't agree with. Let Ritchey (Right the First Time) recently told a bunch of us that he talked with some of the people that wrote these app notes and asked where they get their information from, and the answer was usually that they just copied it from the previous app note, back to the beginning of app notes in some cases. For example, doe sit REALLY make sense to put an inductor in the form of a ferrite bead in series with a power supply when the goal is to lower inductance, not raise it?
There is another myth that got it's genesis in something wrong in an early ECL design guide, and has been with us ever since.
Yes, please do a second one! It's fascinating stuff.
Too short! This could have been 30m longer and I'd still watch it to the end. There is a huge shortage of tech lecturers online and Bil is just a treat to listen to.
More vids like this please!
It's such a treat seeing any of BIls videos, thank you so much for sharing, I look forward to the next one.
Fun stuff! Would love to see more on the C64 reflection issue and how you identified it!
Yeah I am excited now that I have the equipment to see what evil I was burying so many years ago.
Such an awesome video Bil! You're really passionate about sharing knowledge and that's greatly appreciated!
Having migrated to FPGA development for high-speed buses (currently doing a Cyclone V with USB 3 transceiver), its amazing how good looking layouts and routing immediately becomes so critical.
To think you are the main reason I got into computers because of the C64.... quite amazing. Great video. I learned lots. Much to think about.
Suddenly, it made sense. The return path follows the signal path because of capacitance, mostly, but also inductance. It follows the path of least impedance, although it doesn't match the one of least resistance.
Thanks for the detailed video, and for the aha moment, Bill!
Longer more in-depth videos are wayyy better than short, gloss-over videos! Always sad when a video comes to an end. :-(
In-depth definitely yes. But I have some constraints - I don’t start viewing videos that are 5 hours long. In fact, I seldom pick a video that is over 1 hour. Sometimes I have a time slot of just 10 or 15 minutes available and I may pick a video that fits in that window. But generally my observation is that even in-depth videos fit in one our. I mean, some videos seem to have a repetitious narration “be extremely careful” and similar repeated ad nauseum. That is not in-depth as far as I am concerned.
You can always watch part of a longer video, and then come back to it at a later point if you dont have time to watch a longer video start to finish in one go? I have personally done this, no it's not ideal.
Maybe long format videos are not what hackaday want to produce, which I think is a shame when people like Bil have such an amazing amount of knowledge and stories they can share. Sure I can go and do my own research to fill in the gaps, but then you miss out on the personality of the presenter.
Heh, I have access to what the average attention span is on my videos, and it could be said that this wasn't short enough.
BilHerd
As I thought I was expressing, in my case it is more a question of available time windows than attention span. On the other hand, splitting long ones is not really working either. If I close my watching - and my iPad -. Next time I open UA-cam, the interrupted story is nowhere to be seen. It may return to the “Recommended for you” offerings maybe in a week or maybe in a month. I guess I could mark it as one I want to watch later, but I have not got to that kind of habit.
Thanks for the high speed ride on this Mr. Herd. Definitely needs more time on this subject...
it's been a while since I did this stuff in university.
BTW - I picked up one of your C128's recently (a C128D). Looking forward to poking around in it and see how the Z80 and 68xx play with one another.
Cheers,
- Eddy
Great video, thanks! I had my first experience of highish speed layout around 5 years ago when I started using TFTs which necessitated some SDRAM. I read all the app notes with their rules of thumb and came to the conclusion it was impossible if you stuck to the rules and wanted a 4 layer board. But as has happened so many times in my career I did my best and it was totally fine. (Measure twice, cut once.)
I would love to see more of this! Thanks for the excellent video, Bil!
Please make a follow up. Great video
It's coming. :)
Great video, Bil.
So many broken thumbs!
Looking forward to followups.
Oh I always wanted to know about the effect of vias. Turns out (at least done as stubs) they cause a way worse effect than right angles. (well yeah I get that its an apples to oranges comparison but people seem to get a lot more bothered about right angles than via placement in traces).
actually, a via is a 90 degree turn in signal path, it two is 90 degree turns infact if you think about how via is put into a PCB
@@優さん-n7m Yeah, and I guess it would depend heavily on the thickness of the PCB too
Nice to see you put so much effort into these movie’s!!
I never thought that the issues with square-corner traces would come up at so high frequencies, I just assumed that the reflections of the propagating waves would produce EMC problems at any frequency haha! Good to know, because it helps you reduce the PCB size and the traces routing. Great vid!
Somewhere in the Multiverse there's a version of me who finished his Science degree and is going "Duh!" after watching this. My version in this universe is just amazed at all the magic. Great video Bil! Thanks for posting.
I was taught to avoid sharp corners because the etch could undercut the resist at those sharp features, reducing the trace width, sometimes etching through the trace. Still a good reason to avoid sharp corners.
who makes the arbitrary time limit? it takes what it takes, we'll hang in there....
I think its the attention span of the average Hackaday viewer that we most want to accommodate. My editor would probably like it if I made more and shorter videos, :) There is just so much post-editing time involved in making a video that I try and cram as much into one as I can.
WOW! Great Great Video - Very Honest Man to know all these things about PCB! Thank You for this expertise and analysis!
I wish I had your set up! Great video please send more.
Can you explain what exactly the TDR device is showing? The units were milivolts in the left, how does that tell you impedance?
He didn't mention it, it is related to the reflected wave ratio, here is a tektronix link that can help to understand www.google.com/url?sa=t&source=web&rct=j&url=www.tek.com/document/fact-sheet/tdr-impedance-measurements-foundation-signal-integrity&ved=2ahUKEwiIvtX33uXnAhUEnZ4KHS_oDRMQFjADegQIAhAB&usg=AOvVaw2NZJP5Oh8zyAC92obfjuq7
Thanks Bil and Hackaday!
This subject is a bit over my head, although it's one I really want to learn.
I wish you would consider doing more of an abstract overview that covers high speed design 101 from the worst case scenario to radio.
I'm a bit hazy about when and how high speed design comes into play.
I wish you would assume the average hobbyist starts with a single sided home etched design, then graduates to a double sided, then further specialties like impedance controlled copper clad, or board houses. These days, IMO, the path of least resistance is to start with AVR's before radio. I don't know the limitations of each of these applications and will inevitably learn the hard way,.. unless lady luck (google) leads me to the right references. I know Alan W. has a bunch of stuff about doing Manhattan Style boards and dead bug/point to point designs. I just haven't played with much radio stuff yet. I don't know how that fits into the possibilities between single and double sided boards. I know I've taken apart lots of consumer products like TV's and Radios that were built on single sided boards w/jumpers. I don't know if the EEs have Tinkerbell on their shoulder adding pixie dust into their designs or if single sided boards are far more practical than they appear due to all the information people share about more advanced designs. Such info would be extremely helpful to me as a dumb hobbyist blindly wronging my way to right ;)
-Jake
What do you mean by radio stuff?
@@優さん-n7m
In my uneducated sense of board design I think of "radio stuff" as anything around or above the UHF frequency range. I think the biggest hobbyist related application of, and introduction to, such things is amateur radio. I am probably mistaken. I just mentally group all of this in a figurative black box labeled magic... for now.
Jake, There’s more to say on this subject than can fit into a UA-cam comment, but suffice to say you’re actually not far off in saying that it’s ‘magic’ once you get past a certain speeds. In fact one of the most popular books on the subject of high speed design joking calls it “black magic” (www.amazon.com/High-Speed-Digital-Design-Handbook/dp/0133957241). It certainly feels that way to most of us, even if you understand the theory.
One thing that’s helpful to remember is that the rules haven’t changed (the math and physics are the same), but lots of things you can safely ignore at low speeds become headaches and other things that seem significant at low speeds may not be as important. So you’ll need to learn new set of rules of thumb and expectations at those speeds, or better yet will need a strong intuitive understanding of the underlying math and physics. Simply put: the faster the speed, the more the geometry of your traces matter.
If you’re interested in working with Bluetooth, FPGAs, PCIe, WiFi, etc it’s going to be difficult to get good results with home-etched boards and rules of thumb. If you’re working with microcontrollers and I2C and SPI etc over short distances, you won’t need to worry about all of this to get it to work, but if you understand it, it will work better.
If you want to learn more, I the above textbook and Eric Bogatin (mentioned in the video: www.bethesignal.com/bogatin/) are great resources to get started. If that’s more than you want to take on now, I highly recommend ‘The Art of Electronics’, if you haven’t read it yet (www.amazon.com/gp/aw/d/0521809266/). It covers lots of key topics in a very approachable way that will help build you up for more of the high speed theory.
From my perspective, if you’re able to do everything you want to do right now with electronics, don’t worry about it, but if you’re running into limitations with your data speeds or want to do wireless transmission without buying a pre-built module, then start small, make mistakes you can learn from, and find someone smart you who will let you ask dumb questions.
@@michaelkling4994
Thanks for the book references.
I'm not having too many problems yet, but I've only done around a dozen or so prototype designs so far with my biggest being a simple ATmega644 AVR Transistor Tester dev board, so nothing serious yet.
I have The Art of Electronics 2nd edition along with a few other books. I generally like to get an abstract idea of what to be on the lookout for, that's all. I like to build up a map in my mind of how things I don't really understand yet connect to what I'm doing or want to do in the near future.
I'm disabled and have all the time in the world. I'm just really really slow at getting projects done and I don't get out much. So...I'm usually asking lots of dumb questions in YT comments ;)
It's all in good fun.
-Jake
Upcycle Electronics, yeah I totally get what you’re saying about having a mental map. That’s part of why I got my EE degree: I could conceptually understand how you could build a circuit out of wires and chips, and how you could code a piece of software, but things got pretty hazy for me in-between.
What I ended up finding out the hard way, was that since electricity is (usually) invisible, the only way to describe what is happening and how it works is through math. I’m certainly not a math genius, but I found that as a I fought through it, math helped provide the language and framework to open up semiconductors and electromagnetism and all the other concepts I needed to fill in the gaps in my mental model.
One of my biggest breakthroughs was discovering frequency-domain mathematics (eg Fourier transforms). It took some time but once I started conceptualizing all signals as collections of sine waves at different frequencies and amplitudes, it opened up all kinds of things, particularly in filters, signal processing, high speed electronics, and electromagnetism.
If you haven’t gone down the frequency-domain rabbit hole yet, I bet it would be well worth you time to spend some time looking up Fourier series and transforms. The overall concept is more important than the details of the math for day to day use, and it helps bridge the gap between what you’re working with now and higher speed boards and projects.
Best of luck to you :-)
Really great video, would be nice to have maybe a series on this?
Would be nice to see a video that covered possible bus speeds of the classic era of 33 Mhz on up into the Ghz range and sample common speeds along the way (for data transfers). Then do a sweep with signals that simulate actual RF (usually based around sine waves) covering HF -> VHF and maybe UHF.
I was curious to see some effects on the different types of signal because of the potential for ringing (which you did show some ripple from the vias)
Thanks for the great video!
Fascinating stuff. More videos like this please
With the right angle traces, your board actually exaggerates the problem a little bit, because each turn of your square snake allows a portion of the trace to couple to the previous turn. Your typical right angle trace that we don't have to worry about will not have geometry that allows the signal trace to couple to itself like that. This just further proves how pointless it is to remove right angles until you get to the GHz-ish ranges.
I would want to see the right angle thing done with just one or two right angles, where the path then really turns (or is Z shaped). I think in that configuration you might have quite a bit of capacitive coupling between the traces, that is I wouldn't be surprised if the picture looked similar had you left the "right angle ends" off the lines.
:) That's right Maxwell's equations only speak to Electric (E) and Magnetic (B) fields, Current (J), and the speed of light (c). No voltage directly, that's the potential in the E field.
That's what I understand though its always a risk for me to make a statement like that.
@@BilHerd Happy to take the heat for you. Thanks for your good work.
Huh? From the Electric field you can work back to 'voltage' ... (What quantity do you suppose the "Electric Field" is measured by?)
Jumped right in to it... and over WAY too soon. Moar!
Bill thanks! This was awesomely helpful for me!
so those 'ceramic bord modules' found integrated on stuff have a reason. the impedance on those should be quite consistent.
Thanks Bil, this is pretty interesting even for a hobbyists like myself.
Superb. Make it a 10 part series. Did I say superb ?
Why does the TDR pulse in a cable travel more slowly than in a PCB?
The corners example was pretty interesting
A pulser is a piece of equipment that generates very short pulses. Pulsars are highly magnetized rotating neutron stars...
Thank you so much for the video, I had a question then, next year I have to choose my major for college and I was wondering to learn these things what major should I choose, is that electrical engineering! I know that my friend is studying electrical but they do not learn about these things, to learn these should they take their master as well or is there any specific major, like electronics engineering?
The post that accompanies this video can be seen at hackaday.com/2019/01/24/video-putting-high-speed-pcb-design-to-the-test/ There are additional graphics there that support what go flying by in the video.
Re: ground planes.
The graphic you show at 5:12 is even true at hundreds of kHz.
Current has little to do with electron flow which is centimeters per hour, yet that's still how we're all brainwashed in University and by KCL circuit simulators.
Only Faraday's and Maxwell's equations correctly explain current density at frequencies above 100kHz. Good luck with those KCL spice simulators.
Kudos for the graphic at 5:12......I can guarantee you most ee graduates today will never understand that.
Years ago I was given a TDR to test wiring and it made things much easier
Er loaned
Just as my philosophy is "it never hurts to be polite", I think its also fair to say, "it never hurts to avoid 90 degree bends"
Rock with Dave. You could debate this forever.
ah so this is what the other end of the rabbit hole looks like! beaut bud keep up the good work!
new subscriber here
Bil Herd! My favorite computers ever, the C64 and 128, and the Amiga line.
Thanks Bil! More of this please.
Great stuff, I'll take plenty more like it!
What is this cool music in the end.
Weave anisotropy induces mode conversion primarily, due to permittivity variation/intra differential skew. The impedance impact is negligible however, because impedance has more to do with the geometry of the copper conductors. Also, 106 glass isn't used by anyone making a high-speed board. It's horrible. A board house wouldn't even allow it. It would be Megtron 6 within 3313 glass
That was really nice! It would be great if you could talk more about what frequency ranges should be concerned with the different techniques e.g. do right angles matter at 320MHz, 3.2GHz, 10GHz, etc.. also what is required for the really high speed links available in recent Xilinx FPGAs like 10Gb/s, 16Gb/s, 25Gb/s
Right angles to matter since at right angles the track width is larger and that impacts the characteristic imedance. Now the design might work just fine, or this impedance mismatch might ruin the design completely. It depends on the frequency of the signal.
3:29 Pretty much the only PCB design advice they gave us at uni and it was wrong haha.
Awesome content as usual
That was really nice. Thanks for your time.
Mittered bend is the best. Statement, that 90° corner bends are OK for PCB design at high frequencies is simply not true. For frequencies around 80GHz, difference in S21 can be as much as 0.4dB between mittered and 90° corner bend and return loss difference around 14dB! I simulated that in ADS for 101.6um thick laminate with Er=3.66 and for 211um wide microstrip line. I bet that this would translate into real world as well. No RF engineer would ever design with 90° corner bends.
More Bill Herd!
Bil actually. ;)
@@BilHerd My Bad!
Black Magic, this is really cool!
If you like black magic, I would highly recommend the book "High Speed Digital Design: A Handbook of Black Magic" by Howard Johnson & Martin Graham
Or Advanced Black Magic
If you look, that book is on the shelf behind me.
@@BilHerd LOL, sure 'nuff. May I ask what the other two volumes next to it are?
...and any opinions on Ott's title, "Electromagnetic Compatibility Engineering"?
@@Peter_S_ They are 'Noise Reduction Techniques in Electronic Systems' and the Motorola 'MECL System Design Handbook'. He talks about them in his other PCB design video.
Right angles make the trace longer, make the signal travel a few pico seconds slower, look ugly and occupy more space.
This stuff almost killed all the fun of design for me. When you have over 10 thousand digital signals, all with picosecond rise/fall times you spend all your time on routing and analysis instead of design. Automated tools for analysis helped but it still took forever. We lost a few guys just out of frustration. Interesting physics but that can carry your interest only so far.
Thanks for the video but can you please get a better sound setup? The white noise in the background is super annoying. As they say, the cobbler always wears the worst shoes.
I think Rick Hartley said it best "copper traces on a signal layer act as a waveguide for fields traveling in the dielectric"
What a great video thanks so much. Very interesting.
Great video, did feel short!
that weave thing they just don't teach in school. they just took the cop-out method and told to calculate using Kevlar.
Twentyfive schmentyfive. MOAR ViDS LiKE THiS PLZ!
Just use a reflectometer. Will show you all the info with no tinkering.
Excellent video.
Very interesting!
Nince to have Bil back :)
supposed to be 10 minute videos?! I'm thinking they should be 30 mins at the very minimum
It started as 30+ minutes, I cut almost 15 out. Hopefully I can work them into the next video.
It’s still basically magic, we know all about electrodynamics and quantum-electrodynamics to very very high precision, but nobody can actually solve the equations for real world applications, the math is just too complex on the macro scale...
Mr Herd. Thank you :)
Wow how can i give this video more than one like?
Legally change your name or forge the credentials and then create a new google user. If caught the google police access your personal profile and it's not pretty, your GPS stops working, your credit rating tanks, and you can only watch kitten video's on youtube.
@@BilHerd kthx
I watched it on my phone , gave a like.... then my bluetooth tethered tablet allowed me to give a second like !! 😅
Great videos, i would prefer videos like these to be longer. 10 minutes is way to short.
High-speed PCB design: designing while high on caffeine.
Let's think about the public's reaction to this video... Like... WHAT? So many errors in content, labeling, units (both time and frequency) that this video needed MASSIVE edits to make it understandable, let alone usable.
These should be 2-hour lectures not 10 minutes.
Good video. 👏🙌👍👌 😊
my oscilloscope can do all of that ist dpo7104
Testing Myths of High-Speed PCB Design, that is what happens in China, fast design.
Testing Design Myths of High-Speed PCBs. ahhh thats better..
I learned to speak English growing up in Indiana, it plagues me to this day.... y'all.
Very interesting -- thanks! For anyone else who doesn't know, Bill Herd is a design-engineering legend; look at the Wikipedia article for more info: en.wikipedia.org/wiki/Bil_Herd
for demonstration black might not be best.
Who would thumbs-down this but ultracrepidarians.
Anyone else slightly disappointed that this video is not about designing PCBs very quickly?
Great video!
FYI "High Speed Design" (for PCB's) has historically referred to routing for high frequency on PCB's. You could search for "Fast Turnaround" design or "Fast Layout Techniques" if you are looking for that kind of info.
Cheers,
- Eddy
Deathlok67 Thanks! MickMake just did a great series on this that I just watched, so it was on my mind. Cheers
You can design complex PCBs well or you can design them fast. But not both, unless you've done 100's of boards.
nice shop
Very interesting.
I too am a "just run a wire over it" kind of guy
Amazing.
What matters more than any of this is the ground under these traces. Zero discussion in this video. What is he testing? Can't be solid ground plane he has vertical trace on backside. No ground at all? A ring around the edges grounding the SMA? Pretty poor. I'd rather see video showing gaps in ground plane and cross wires underneath.
It is a four layer board, so the inner two layers are probably a solid ground plane
14:18 Touch screen lol
Oh by High Speed you ment frequency. I thought it was something like how to design a PCB in 20 min.
Mind bent.
ha ha ha
I thought the video would be about quickly making a PCB from circuit design to completed board in just a few days. Disappointed now. Unsubscribing! /sarc :-o
thanks for doing this